A Rapid Food Screener to Assess

Fat and Fruit and Vegetable Intake

Gladys Block, MD, PhD, Christina Gillespie, MD, Ernest H. Rosenbaum, MD, Christopher Jenson, PhD, RD

Background: The U.S. Preventive Services Task Force recommends that Americans lower dietary fat and
cholesterol intake and increase fiber and fruit/vegetables to reduce prevalence of heart
disease, cancer, stroke, hypertension, obesity, and non–insulin-dependent diabetes melli-
tus in the United States. To provide preventive services to all, a rapid, inexpensive, and
valid method of assessing dietary intake is needed.
Methods: We used a one-page food intake screener based on national nutrition data. Respondents
can complete and score the screener in a few minutes and can receive immediate, brief
feedback. Two hundred adults self-administered the food screener. We compared fat, fiber,
and fruit/vegetable intake estimates derived from the screener with estimates from a
full-length, 100-item validated questionnaire.
Results: The screener was effective in identifying persons with high-fat intake, or low–fruit/
vegetable intake. We found correlations of 0.6 – 0.7 (p⬍0.0001) for total fat, saturated fat,
cholesterol, and fruit/vegetable intake. The screener could identify persons with high
percentages of calories from fat, total fat, saturated fat, or cholesterol, and persons with low
intakes of vitamin C, fiber, or potassium.
Conclusions: This screener is a useful tool for quickly monitoring patients’ diets. The health care
provider can use it as a prelude to brief counseling or as the first stage of triage. Persons
who score poorly can be referred for more extensive evaluation by low-cost paper-and-
pencil methods. Those who still have poor scores at the second stage ultimately can be
referred for in-person counseling .
Medical Subject Headings (MeSH): diet, nutrition, primary prevention, screening (Am J
Prev Med 2000;18(4):284 –288) © 2000 American Journal of Preventive Medicine

T
he report of the U.S. Preventive Services Task lack the skills to obtain a thorough dietary history . . .
Force recommended that Americans lower their and to offer specific guidance.”
fat and cholesterol intake, and increase fiber To achieve the dietary recommendations of the Task
intake to reduce the prevalence of heart disease, can- Force, providers need simple, inexpensive, rapid, and
cer, stroke, hypertension, obesity, and non–insulin- valid tools that can provide a snapshot of the patient’s
dependent diabetes mellitus in the United States pop- diet, provide immediate feedback to the patient, and
ulation.1 The report went on to state that “Although afford the clinician the opportunity to use the influ-
immunizations and screening tests remain important ence of the office to reinforce the feedback. We report
preventive services, the most promising role for prevention in the testing of such an instrument.
current medical practice may lie in changing the personal A brief fat screener, developed earlier,2 required
health behaviors of patients long before clinical disease devel- computerized scoring. More recently, a one-page Food
ops.” The task force acknowledges, however, that clini- Screener has been developed, designed to obtain a
cians often fail to provide preventive services such as valid measure of nutrient intake in much less time than
nutritional screening and counseling, due, in part, to required by traditional food records or extensive food-
insufficient time with patients and the fact that “many frequency questionnaires. The screener can be com-
pleted and scored in 5 minutes or less, and does not
require difficult calculations or computer analysis. It
From the School of Public Health, University of California (Block), includes the top sources of fat and of fruits and
Berkeley, California; Tufts University, School of Medicine (Gillespie),
Boston, Massachusetts; UCSF Mount Zion Cancer Center, Medical vegetables in the diets of Americans, as determined by
Oncology and Hematology (Rosenbaum), San Francisco, California; national surveys and recent research.
and Shaklee Corporation (Jenson), San Francisco, California In a multi-ethnic population in the San Francisco Bay
Address correspondence and reprint requests to: Gladys Block,
PhD, 426 Warren Hall, School of Public Health, University of area, we examined agreement between the Food
California, Berkeley, CA 94720. E-mail [email protected]. Screener scores and nutrient estimates produced by the

284 Am J Prev Med 2000;18(4) 0749-3797/00/$–see front matter

© 2000 American Journal of Preventive Medicine • Published by Elsevier Science Inc. PII S0749-3797(00)00119-7
Table 1. Demographics of study population (n ⫽ 208) full-length, eight-page Food Frequency Questionnaire. Pa-
tients self-administered both forms. As incentive, we offered
Demographic Frequency Percent an individual nutritional analysis, the results of which were
Gender kept confidential from company management. Of the 402
Male 74 35.6 employees invited, 208 chose to participate in this program.
Female 134 64.4 Females comprised almost two thirds of the study popula-
Race tion (Table 1). The largest ethnic group was white, constitut-
Hispanic 15 7.2 ing 65% of the sample. Ages ranged from 20 to 69 years, but
White, not Hispanic 135 64.9 more than one half of the respondents were between the ages
African American 7 3.4 of 30 and 49 years.
Asian, Pacific Islander 45 21.6
The Food Screener is a one-page self-administered tool, in
Missing/not specified 6 2.9
Age two sections. The Meat/Snacks section is comprised of 15
20–29 21 10.1 items designed to capture dietary fats. The Fruit/Vegetable
30–39 58 27.9 section is comprised of seven items designed to capture fruit
40–49 68 32.7 and vegetable intake, fiber, and micronutrients found in
50–59 40 19.2 fruits and vegetables. (See Table 2 footnote for items and
60–69 13 6.3 scoring.) The Food Screener requires no software for inter-
70 or older 0 0 pretation because its purpose is to give immediate feedback
Missing 8 3.8
to respondents. Thus the participants themselves score the
screener.
The “gold standard” against which we compared the
Block full-length Food Frequency Questionnaire. We screener was the 1995 Block 100-item Food Frequency Ques-
examined the role of race, gender, and age as covari- tionnaire.3 Several published studies have validated the Food
ates, and constructed prediction equations based on Frequency Questionnaire as an accurate tool for measuring
Food Screener scores. dietary nutrient intake.3–5 It gives nutrient estimates that
correlate well with those obtained by much more extensive
methods of measuring food intake, such as multi-day food-
Methods intake records. We analyzed the full-length Food Frequency
We invited employees of a company in the San Francisco Bay Questionnaire using DIETSYS, a program developed by the
area to participate in a Nutrition Education and Research National Cancer Institute to calculate individuals’ daily nutri-
Program, conducted by the University of California. Each ent intake based on responses to the questionnaire. To
employee received the one-page Food Screener and the calculate nutrient intake using the full-length questionnaire,

Table 2. Correlation between food screener scores and full-length questionnaire nutrient estimates (n⫽208)
Spearman Significance
r Value (p Value)

Meat/Snack Screener Scorea—correlation with the following

nutrient estimates from full-length questionnaire:
Total fat (gms per day) 0.69 0.0001
Saturated fat (gms per day) 0.72 0.0001
Monounsaturated fat (gms per day) 0.67 0.0001
Dietary cholesterol (mg per day) 0.60 0.0001
Percent fat (daily percent of total calories) 0.63 0.0001
Fruit/Vegetable Screener Scoreb—(excluding
beans/legumes) correlation with the following nutrient
estimates from full-length questionnaire:
Fruit/vegetable servingsc 0.71 0.0001
Vitamin C (mg per day) 0.57 0.0001
Magnesium (mg per day) 0.41 0.0001
Dietary fiber (gms per day) 0.50 0.0001
Potassium (mg per day) 0.49 0.0001
Fruit/Vegetable Screener Scoreb—(beans/legumes
included) correlation with the following nutrient estimates
from full-length questionnaire:
Magnesium (mg per day) 0.46 0.0001
Dietary fiber (gms per day) 0.62 0.0001
Potassium (mg per day) 0.52 0.0001
a
Meat/Snacks: 15 items (meats, dairy, spread, snacks). Response categories: once per month or less, 2–3 times per month, 1–2 times per week,
3– 4 times per week, or 5 or more times per week. Those categories were scored 0, 1, 2, 3, or 4. Scores range 0 – 60.
b
Fruit/Vegetables: 7 items (fruits, vegetables, juices). Response categories: less than once per week, about once per week, 2–3 times per week,
4 – 6 times per week, every day, and 2 or more times per day; scored 0, 1, 2, 3, 4, 5. Scores range 0 –35.
c
Estimates from the full-length questionnaire were calculated using USDA Food Pyramid definitions of servings.

Am J Prev Med 2000;18(4) 285

we multiplied the reported frequency of consumption of each by the full-length questionnaire was set at five or more
food by the nutrient content of that food, and by the reported servings, consistent with several national guidelines.
portion size (small, medium, or large). The software then “Quite low” was defined as three to four servings, and
uses age- and gender-specific portion sizes for small, medium, “very low” fruit and vegetable consumption was defined
and large that are derived from national data.3
as one or two servings per day. Thirty-seven of the 56
We conducted the statistical analyses using SAS V6.11. To
examine the predictive value of the screener scores in iden-
persons with low scores (⬍11) on the fruit/vegetable
tifying persons for whom dietary counseling might be bene- screener were indeed “very low” by the gold standard
ficial, we dichotomized the Fruit/Vegetable and Meats/ (66%), and an additional 13 were “quite low.” Thus,
Snacks scores at the lower and upper quartile of their 89% of persons who were low on the fruit/vegetable
respective distributions. To assess the validity of the Food screener were indeed “very low” or “quite low” by the
Screener, we calculated Spearman rank-order correlation gold standard.
coefficients between Food Screener scores and nutrient val- “Very high” fat intake was defined as 35% or more of
ues measured by the full-length Food Frequency Question- energy from fat (Table 3). Of those who scored 23 or
naire. We repeated these analyses for whites and for “other higher on the screener, 57% were indeed “very high”
than white.” Study participants who did not indicate their according to the full-length questionnaire, and 88%
race were excluded from this part of the analysis. We used
were above the desirable 30% of energy. Thus, all but
linear regression models to develop prediction equations,
which could be used to estimate actual nutrient intake based
12% of the persons identified by the screener as having
on Food Screener scores. Age, gender, and race were consid- elevated fat intake could benefit from advice to reduce
ered as potential covariates and, where appropriate, included their fat intake. For saturated fat intake, 63% of those
in the prediction equations. Extreme outliers were excluded flagged by the screener were very high, and another
from the linear regression analyses (but not from the main 29% were quite high in saturated fat intake as mea-
correlation analysis) to allow the prediction equations to sured by the full-length questionnaire (Table 3).
estimate nutrient intake more accurately for the majority of We used linear regression techniques to construct
respondents. prediction equations for each nutrient on the
Food Screener. These are reported in Table
Results 4, and can be used to translate the Food
See Screener scores into estimates of daily nutri-
The Food Screener ranked subjects similarly related ent intake. We examined age, gender, and
to estimates from the Block full-length Food Commentary race as potential covariates in the regression
Frequency Questionnaire, used here as the on page 354. analysis. These variables were included in the
gold standard, for a number of important prediction equations when their level of sig-
nutrients (Table 2). Spearman rank-order nificance was found to be p⬍0.05. Gender
correlation coefficient (r⬎0.60) showed that was a significant covariate for all nutrients on
the Food Screener ranked subjects quite well the Food Screener. Age was a significant
with respect to dietary intake of total fat, saturated fat, covariate for vitamin C, potassium, and magnesium;
dietary cholesterol, and percent of calories from fat. We and race was significant for only dietary cholesterol
also obtained an excellent correlation of the screener (Table 4).
with servings of fruits and vegetables (r⫽0.71), and with
vitamin C as estimated by the full-length questionnaire.
Further analysis indicated that inclusion of the bean/ Discussion
legume item on the screener produced an improved The correlations of the meats/snacks score with fats
correlation of the screener score with the more exten- and cholesterol indicate that the screener can provide
sive questionnaire’s estimates of magnesium, potas- estimates of these nutrients similar to rankings ob-
sium, and fiber (Table 2). tained from a full-length nutrition questionnaire. If
Data analyses restricted to the white study subjects intake of fruits and vegetables is of interest, the fruit/
(n⫽135) produced similar results. All nutrient correla- vegetable screener (with the omission of the beans
tions remained roughly the same, with a slight increase item) produced an excellent estimate of servings of
in the vitamin C correlation coefficient (data not those foods. The correlations of the fruit/vegetables
shown). The analyses performed on the data from the score with associated nutrients as opposed to foods
“other than white” study group (n⫽67) produced re- tended to be somewhat lower, due in part to the fact
sults similar to those of the entire population for all that these nutrients (e.g., vitamin C) are also obtained
nutrients, with a slight decrease in the vitamin C from fortified foods, such as breakfast cereals, found on
correlation coefficient (data not shown). All results the full-length questionnaire but not on the screener.
were statistically significant, with p values less than 0.01. Thus, the correlations seen here indicate that the
The predictive value of the screener scores was good screener can provide a reasonable picture of respon-
(Table 3). For purposes of this analysis, the upper dents’ intake of fat, and of fruits and vegetables.
category of fruit/vegetable consumption as determined The health care provider can use the screeners as a

286 American Journal of Preventive Medicine, Volume 18, Number 4

Table 3. Predictive value of screener scores
Servings by full-length questionnaire
a
Fruit/Vegetable screener 1–2 3–4 5ⴙ Total

Fruit/vegetable (FV) screener score

Low (⬍11) 37 13 6 56
Not low 34 73 48 155
Total 71 86 54 211

Percent of energy from fat by full-length questionnaire

b
Fat Screener Less than 30% 30–34% 35% or higher Total

Meats/Snacks screener score

High (23⫹) 6 16 29 51
Not high 78 47 32 157
Total 84 63 61 208

Saturated fat intake by full-length questionnaire

Fat Screenerc Lower quartile Middle Upper quartile Total
(⬍15 g) (15–26 g) (ⱖ27 g)

Meat/Snacks screener score

High (23⫹) 4 15 32 51
Not high 47 89 21 157
Total 51 104 53 208
a
Chi-squared p ⬍ 0.001. Sensitivity (ability to detect low intake) 37 ⫼ 71 ⫽ 52%.
Specificity (ability to rule out people with high intake) 48 ⫼ 54 ⫽ 86%.
Predictive value of a positive (low FV) score 37 ⫼ 56 ⫽ 66%.
b
Chi-squared p ⬍ 0.001. Sensitivity (ability to detect high-fat intake) 29 ⫼ 61 ⫽ 52%.
Specificity (ability to rule out low-fat intake) 78 ⫼ 84 ⫽ 93%.
Predictive value of a positive (high-fat) score: 29 ⫼ 51 ⫽ 57%.
c
Chi-squared p ⬍ 0.001. Sensitivity (ability to detect high intake) 32 ⫼ 53 ⫽ 60%.
Specificity (ability to rule out low intake) 47 ⫼ 51 ⫽ 87%.
Predictive value of a positive (high-fat) score 32 ⫼ 51 ⫽ 63%.

Table 4. Prediction equations for daily nutrient intake

based on food screener scoresa
stimulus for delivery of brief nutrition advice. The
screener identifies those in need of advice, and pro- Meat/Snack Equations:
vides self-scoring that informs the user of his or her Total fat (gms) ⫽ 32.7 ⫹ 2.4 (Meat/Snack score) ⫹ 11.2S
status. The health care provider can use the authority of Saturated fat (gms) ⫽ 9.4 ⫹ 0.88 (Meat/Snack score) ⫺3.5S
the profession to reiterate those points, and could Percent fatb ⫽ 19.8 ⫹ 0.6 (Meat/Snack score) ⫹ 2.3S
point to foods frequently eaten by the respondent as Dietary cholesterol (gms) ⫽ 120 ⫹ 7.8 (Meat/Snack score)
first targets for dietary improvement. In addition, pro- ⫺ 54.65S ⫹ 36.6R
viders can use the screener as first-stage triage; patients Fruit/Vegetable Equations:c
who score poorly on fat or fruit/vegetable intake could Fruit/vegetable servings (Pyramid definitions of servings
take the full-length instrument or receive more exten- per day) ⫽ ⫺0.23 ⫹ 0.37 (Fruit/Vegetable score) ⫺0.55S
Vitamin C (mg) ⫽ 56.5 ⫹ 6.6 (Fruit/Veg/Beans score
sive assessment and counseling by a nutrition ⫺ 26.7S ⫺0.45A
professional. Magnesium (mg) ⫽ 272 ⫹ 11.6 (Fruit/Veg/Beans score)
The effectiveness of the Food Screener is based on ⫺92.3S ⫺1.7A
the fact that it includes the most important sources of Dietary fiber (gms) ⫽ 7.9 ⫹ 0.74 (Fruit/Veg/Beans score)
these nutrients in the diets of most Americans. Corre- ⫺4.5S
Potassium (mg) ⫽ 2348 ⫹ 114.8 (Fruit/Veg/Beans score)
lation coefficients differed only slightly between the ⫺759S ⫺13.8A
“white” and “other than white” categories. The screener a
Variables are defined as follows:
may be less appropriate for persons with unusual S ⫽ Sex: Male ⫽ 0, Female ⫽ 1
dietary practices, such as recent immigrants. R ⫽ Race: White ⫽ 0, Nonwhite ⫽ 1
Clearly various subcomponents of total fat (e.g., A ⫽ Age: Age can be directly substituted for the A variable.
b
Accuracy of the percent fat prediction will depend in part on the
saturated fat, monounsaturated fat) may have different extent of empty calories (such as from soft drinks or alcoholic
health effects.1 Saturated fat increases serum lipids, and beverages) in the respondent’s diet.
c
as shown in Table 2, the screener does a good job of If only servings of fruits and vegetables is of interest, use the screener
without beans, and use the equation for servings per day. If other
identifying individuals with high saturated fat intake nutrients such as fiber are of interest, use the screener including
and, in fact, correlates more highly with that than with beans, and use the other equations.

Am J Prev Med 2000;18(4) 287

total fat. On the other hand, a high monounsaturated Questionnaire were both frequency-type instruments
fat intake may reduce the risk of some diseases, and (but with different scoring systems), and were self-
Table 2 also reveals a high correlation with monoun- administered at approximately the same time, these
saturated fat. This illustrates the fact that in the United correlations probably overestimate the correlation of
States, almost all foods high in monounsaturated fat are the screeners with “truth.” On the other hand, these
also high in saturated fat. In data from the Third screeners have also been compared with data from
National Health and Nutrition Examination Survey, detailed 4-day food records (A. Dowdy et al., unpub-
seven of the top ten foods contributing saturated fat are lished observations, 1994). Those results were also very
also among the top ten monounsaturated fat contribu- good, indicating that the screener ranked people sim-
tors (GB, unpublished observations). If a respondent ilarly to much more detailed and extensive dietary
scores high on this screener, he or she almost certainly methods.
has a high saturated fat intake, and high monounsatu- The Food Screener used in this study was developed
rated fat intake. Physicians and patients who wish to to assess fat, fiber, and fruit and vegetable intake. These
improve their lipid profile would first need to achieve a nutrients are most closely associated with morbidity and
lower score on this screener, and then address increas- mortality, and thus are of interest to physicians, epide-
ing monounsaturated fat from good sources that do not miologists, nutritionists, and diet-conscious individuals.
also have a high saturated fat content, such as olive oil. The screeners provide a reasonably accurate ranking of
The meat/snacks and fruit/vegetable scores were nutrient intake, similar to that of a full-length dietary
designed simply to rank individuals from low to high, questionnaire. Thus, they help identify persons with
and to provide immediate feedback based on those high-fat or high-cholesterol intake, or with low fruit and
scores; use of the prediction equations is not necessary. vegetable intake. Because the screeners are brief, and
However, the prediction equations may be useful for can be self-administered and self-scored, they offer an
estimating intake of some nutrients. To use the predic- inexpensive way to provide basic dietary evaluation and
tion equations properly, the investigator or physician feedback to all patients.
should obtain information regarding the individual’s
age, sex, and ethnic group. Although both the screener This work was supported in part by the Harold Dobbs Cancer
score and the prediction equations can provide reason- Research Fund, UCSF/Mt. Zion Cancer Center, and the
able estimates of nutrient intake, misclassification of Shaklee Corporation.
some individuals with regard to their nutrient intake is
inevitable. For screening and counseling, the analyses
in Table 3 indicate that the great majority of persons References
flagged by the screeners could indeed benefit from 1. U.S. Preventive Services Task Force. Guide to clinical preventive services,
dietary change. For nutrition research, full-length in- 2nd ed. Alexandria, VA: International Medical Publishing, 1996.
2. Block G, Clifford C, Naughton MD, Henderson M, McAdams M. A brief
struments are to be preferred. dietary screen for high fat intake. J Nutr Educ 1989;21:199 –207.
The gold standard used in this study, the 100-item 3. Block G, Hartman AM, Dresser CM, Carroll MD, Gannon J, Gardner L. A
Block dietary questionnaire, has itself been validated data-based approach to diet questionnaire design and testing. Am J
Epidemiol 1986;124:453– 69.
against much more intensive assessment methods,
4. Block G, Woods M, Potosky A, Clifford C. Validation of a self-administered
namely 12 to 16 days of diet records, and shown to diet history questionnaire using multiple diet records. J Clin Epidemiol
produce good nutrient estimates and correlations.4,5 1990;43:1327–35.
The present research compared the Food Screener 5. Mares-Perlman JA, Klein BEK, Klein R, Ritter LL, Fisher MR, Freudenheim
JL. A diet history questionnaire ranks nutrient intakes in middle-aged and
with this full-length questionnaire. It should be noted older men and women similarly to multiple food records. J Nutr 1993;123:
that since the screener and criterion Food Frequency 489 –501.

288 American Journal of Preventive Medicine, Volume 18, Number 4